This position is for building, deploying and evaluating the institute's Arabic AI tutor for mathematics and computer science: preparing Arabic training data, fine-tuning, aligning and evaluating LLMs on Shaheen III and Ibex, and integrating the tutor into the learning platform for a pilot in at least 10 Saudi schools.

• Collect and prepare Arabic training data: Ministry of Education textbooks, exams and curriculum standards, and consented teacher–student chats from the platform.

• Build an Arabic mathematics and computer science test set and benchmark commercial and open Arabic LLMs on it.

• Fine-tune a commercial LLM and an open Arabic model (such as Jais) on Shaheen III and Ibex.

• Align the tutor with teachers’ rankings of its replies using reward models and preference training (RLHF or DPO).

• Build guardrails for school students: no answer giving, correct mathematics, level-appropriate language and cultural fit.

• Integrate the tutor into the learning platform, with Arabic (right-to-left) student and teacher interfaces.

• Build the tools for consented chat recording, teacher ranking of tutor replies and short surveys after each chat.

• Serve the models for real-time chat (100 ms per token or less) and keep the service reliable during the school year.

• Roll the platform out to at least 10 schools and build the tools that evaluate the pilot.

• Protect student data: parent consent, access control and the Saudi Personal Data Protection Law.

• Contribute to papers, invention disclosures and RDIA progress reports.

• Advanced proficiency in Python and PyTorch

• LLM fine-tuning with Hugging Face Transformers, PEFT and TRL

• Distributed training on SLURM-based GPU clusters (DeepSpeed, FSDP)

• Reward modeling and preference optimization (PPO, DPO)

• LLM evaluation, including human rating and LLM-as-judge

• Text data pipelines and Arabic NLP, including Arabic OCR

• Web back ends and LLM APIs: streaming chat, prompt management and logging

• Model serving and inference optimization (vLLM, quantization)

• Docker, CI/CD, Git and deployment on cloud or on-premises servers

• Experience building right-to-left (Arabic) interfaces is preferred

• Knowledge of the Saudi Personal Data Protection Law is preferred

• MSc in Computer Science, Artificial Intelligence, or a related field, or a BSc with at least 1 years of relevant experience

• Hands-on experience fine-tuning and evaluating large language models

• Experience building and deploying AI applications used by real users

• Professional proficiency in Arabic and English

• Ability to work with teachers and school staff and to travel to schools in the Kingdom

• Experience with preference optimization (RLHF, DPO) is preferred

• Experience in education technology, or in teaching mathematics or programming, is preferred